Machine Learning Uncovers Key Risk Factors Behind Dementia in Parkinson’s Patients

In a major step forward for Parkinson’s care, researchers have used machine learning and UK Biobank data to predict who is most at risk of developing Parkinson’s disease dementia (PDD), highlighting genetics and treatable conditions like hypertension and diabetes as key factors.

Study: Predicting dementia in people with Parkinson’s disease. Image Credit: LightField Studios/Shutterstock.com

Published in Nature, the study validated its findings using data from the Parkinson’s Progression Markers Initiative (PPMI), applying Shapley additive explanation (SHAP) analysis and Mendelian randomization to show that managing blood pressure and glucose levels could help reduce dementia risk in people with Parkinson’s.

Background

Parkinson’s disease (PD) is the second most common neurodegenerative condition, and it frequently leads to cognitive decline—about 45 % of patients go on to develop Parkinson’s disease dementia (PDD) within 10 years.

While both genetic and environmental factors (including pesticide exposure and chronic conditions like diabetes) have been linked to this progression, existing predictive models are limited. Many rely on narrow datasets or focus only on motor symptoms or genetic profiles, overlooking the broader range of clinical and lifestyle data collected in everyday healthcare.

This study aimed to close that gap. Using the UK Biobank’s extensive data—including genetics, comorbidities, and lifestyle factors—researchers built machine learning models to predict PDD risk. They paired this with explainable AI tools and Bayesian networks to understand how different risk factors interact. Mendelian randomization was then used to test whether some of those factors play a direct, causal role in dementia development.

Study Design and Methods

The study drew from two major sources: the UK Biobank and the PPMI. The UK Biobank offered cross-sectional data from over 500,000 participants, capturing everything from genetic profiles to lifestyle habits. The PPMI added longitudinal data on 1500 Parkinson’s patients, allowing for validation over time.

From these sources, researchers identified 3541 PD and 487 PDD cases in the UK Biobank, and 637 PD cases (122 with cognitive impairment) in the PPMI. Genetic analysis zeroed in on single-nucleotide polymorphisms (SNPs) linked to Parkinson’s and dementia, with additional filtering to ensure genetic independence.

They trained several machine learning models—gradient boosting, random forests, and logistic regression—using nested cross-validation. Predictive performance was measured using area under the curve (AUC), while SHAP values highlighted the most influential predictors. Bayesian networks were used to uncover how genetic, clinical, and environmental variables interact.

To determine causality, Mendelian randomization tested whether conditions like hypertension and type 2 diabetes directly influence PDD risk. Genetic instruments were drawn from GWAS datasets, and results were confirmed with sensitivity analyses to control for pleiotropy.

Key Findings and Analysis

PDD patients in both cohorts tended to be older men (UKB: 69.8 % male, mean age 64.1; PPMI: 61.5 % male, mean age 67.1). Among the machine learning models, random forests performed best, with AUC values ranging from 0.61 to 0.65—moderate accuracy, but meaningful for risk stratification.

Genetic data contributed the largest share of predictive power (49.3 %), led by the polygenic risk score PGS4281 and SNPs rs769449 and rs6859. Demographics (24.3 %) and comorbidities (15.7 %) followed, with age, depression, and male sex associated with higher PDD risk.

Interestingly, a higher body mass index (BMI) appeared to offer a protective effect.

Bayesian networks revealed intuitive links—such as how certain genetic markers align with comorbid conditions—and Mendelian randomization confirmed that both hypertension and type 2 diabetes have a causal relationship with increased PDD risk. These associations remained strong even after controlling for confounding variables. Lifestyle factors like smoking and obesity, however, did not show a causal connection in this analysis.

The overall message was that while genetics set the stage, modifiable conditions like hypertension and diabetes offer potential levers for intervention.

Insights and Discussion

The most influential predictors were genetic, especially the dementia-linked polygenic risk score PGS4281 and APOE-related SNPs. Demographic factors (age, sex) and comorbidities (depression, hypertension) followed closely. Higher BMI emerged as a surprising protective factor, and novel gene-environment links were uncovered, such as an association between SNCA variants and BMI.

While the study has its limitations, including potential bias from ICD coding and the demographic makeup of the cohorts, it offers a more actionable understanding of who is most at risk. Validating the results in the PPMI cohort strengthens confidence in their broader applicability.

The biggest takeaway from this study is that managing blood pressure and glucose levels in Parkinson’s patients could help delay or reduce the likelihood of dementia—an insight with direct clinical value.

Conclusion

This research marks a significant advance in identifying and managing dementia risk in Parkinson’s patients. While genetic predisposition remains the strongest predictor, modifiable conditions like hypertension and type 2 diabetes also play a clear causal role. The machine learning models showed moderate predictive accuracy (AUC ~0.62), with genetics accounting for nearly half of the predictive power.

By combining multimodal data with causal inference methods, the study provides a foundation for personalized risk assessment and opens new pathways for early intervention and prevention.

Journal Reference

Aborageh, M., Hähnel, T., Martins Conde, P. et al. Predicting dementia in people with Parkinson’s disease. npj Parkinsons Dis. 11, 126 (2025). DOI:10.1038/s41531-025-00983-4. https://www.nature.com/articles/s41531-025-00983-4

Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.

Citations

Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Nandi, Soham. (2025, May 20). Machine Learning Uncovers Key Risk Factors Behind Dementia in Parkinson’s Patients. AZoRobotics. Retrieved on May 20, 2025 from https://www.azorobotics.com/News.aspx?newsID=15966.

  • MLA

    Nandi, Soham. "Machine Learning Uncovers Key Risk Factors Behind Dementia in Parkinson’s Patients". AZoRobotics. 20 May 2025. <https://www.azorobotics.com/News.aspx?newsID=15966>.

  • Chicago

    Nandi, Soham. "Machine Learning Uncovers Key Risk Factors Behind Dementia in Parkinson’s Patients". AZoRobotics. https://www.azorobotics.com/News.aspx?newsID=15966. (accessed May 20, 2025).

  • Harvard

    Nandi, Soham. 2025. Machine Learning Uncovers Key Risk Factors Behind Dementia in Parkinson’s Patients. AZoRobotics, viewed 20 May 2025, https://www.azorobotics.com/News.aspx?newsID=15966.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type
Submit

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.